Remote Sensing of Environment 114 (2010) 286–298

Contents lists available at ScienceDirect

Remote Sensing of Environment

j ourna l homepage: www.e lsev ie r.com/ locate / rse

Detecting water stress effects on fruit quality in orchards with time-series PRIairborne imagery

L. Suárez a, P.J. Zarco-Tejada a,⁎, V. González-Dugo a, J.A.J. Berni a, R. Sagardoy b, F. Morales b, E. Fereres a,c

a Instituto de Agricultura Sostenible (IAS), Consejo Superior de Investigaciones Científicas (CSIC), Alameda del Obispo, s/n, 14004, Córdoba, Spainb Department of Plant Nutrition, Experimental Station of Aula Dei, CSIC, Apdo 13034, 50080, Zaragoza, Spainc Department of Agronomy, University of Cordoba, Campus Universitario de Rabanales, 14014, Córdoba, Spain

⁎ Corresponding author. Tel.: +34 957 499 280, +34 6252.

E-mail address: pzarco@ias.csic.es (P.J. Zarco-TejadaURL: http://quantalab.ias.csic.es (P.J. Zarco-Tejada).

0034-4257/$ – see front matter © 2009 Elsevier Inc. Aldoi:10.1016/j.rse.2009.09.006

a b s t r a c t

a r t i c l e i n f o

Article history:Received 16 June 2009Received in revised form 8 September 2009Accepted 12 September 2009

Keywords:Photochemical Reflectance Index (PRI)EPSWater stressFruit qualityTSS/TAMultispectral remote sensingThermal

Amethodology for the assessment of fruit quality in crops subjected to different irrigation regimes is presented.Highspatial resolutionmultispectral and thermal airborne imagerywere used tomonitor crown temperature andthe Photochemical Reflectance Index (PRI) over three commercial orchards comprising peach, nectarine andorange fruit trees during 2008. Irrigation regimes included sustained and regulated deficit irrigation strategies,leading to high variability of fruit quality at harvest. Stem water potential was used to monitor individual treewater status on each study site. Leaf samples were collected for destructive sampling of xanthophyll pigments toassess the relationship between the xanthophyll epoxidation state (EPS) and PRI at leaf and airborne-canopylevel. At harvest, fruit size, Total Soluble Solids (TSS) and Tritatable Acidity (TA) were measured to characterizefruit quality. A statistically significant relationship between EPS and PRI was found at the leaf (r2=0.81) andcanopy level (r2=0.41). Airborne-derived crown PRI calculated from the imagery acquired during the fruitgrowth was related to the ratio of the total soluble solids normalized by the tritatable acidity (TSS/TA), anindicator of fruit quality measured on the same trees, yielding a coefficient of determination of r2=0.50. Therelationship between the integral of PRI time-series and TSS/TAyielded a coefficient of determination of r2=0.72(peach) and r2=0.61 (nectarines). On the contrary, the relation between TSS/TA and the time-series of crownthermal imagery was very weak (r2=0.21 and 0.25 respectively). These results suggest that a physiologicalremote sensing indicator related to photosynthesis, such as PRI, is more appropriate for fruit quality assessmentthan crown temperature, the established method of water stress detection, which is more related to crowntranspiration. A radiative transfer modelling study was conducted to assess the potential validity of thismethodology for fruit quality assessment when using medium spatial resolution imagery. The analysis showsimportant effects of soil and shadows on the PRI vs EPS relationship used for fruit quality assessment if non-purecrown reflectance was extracted from the imagery.

76 954 937; fax: +34 957 499

).

l rights reserved.

© 2009 Elsevier Inc. All rights reserved.

1. Introduction

Twenty-five years ago, thermal information was chosen for theremote sensing of water stress in crops (Jackson et al., 1981; Idso,1982a,b) because the spectral vegetation indices that existed at thattime were not nearly as sensitive to water deficits as those derivedfrom canopy temperature (Jackson et al., 1983). Thermal remotesensing of water stress was first performed using spectrometers atground level (Idso et al., 1981; Jackson et al., 1977, 1981), but otherapproaches have been developed more recently. These included theuse of airborne thermal imagery (Cohen et al., 2005; Leinonen & Jones,2004; Sepulcre-Cantó et al., 2007) and satellite thermal information incombination with 3D radiative transfer models to understand the

effects of scene thermal components on large ASTER pixels (Sepulcre-Cantó et al., 2009). Notwithstanding the advances in thermaldetection, the visible part of the spectrum has also been useful forpre-visual water stress detection based on indices that use bandslocated at specific wavelengths where photosynthetic pigments areaffected by stress condition. This is the case of the PhotochemicalReflectance Index (PRI) (Gamon et al., 1992) that has been proposedto assess vegetation water stress based on xanthophyll compositionchanges (Peguero-Pina et al., 2008; Suárez et al., 2008, 2009; Thenotet al., 2002). The PRI was presented as an indicator of the epoxidationstate of the xanthophylls pool or, what is the same, the proportion ofviolaxanthin that has been converted into zeaxanthin under stressconditions (Gamon et al., 1992). For water stress detection, PRI couldbe an alternative to thermal remote sensing, enabling the use of low-cost imaging sensors with high spatial resolution capabilities that arenot possible in the thermal domain (Suárez et al., 2008, 2009).

In addition, the PRI is an index that was first formulated as anindicator of photosynthetic efficiency, but is also an indicator of

287L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

photosynthesis rate through light use efficiency (Asner et al., 2005;Drolet et al., 2005; Fuentes et al., 2006; Guo & Trotter, 2004; Nakajiet al., 2006; Nichol et al., 2000, 2002; Serrano & Peñuelas, 2005; Simset al., 2006; Strachan et al., 2002; Trotter et al., 2002) and throughchlorophyll fluorescence (Dobrowsky et al., 2005; Evain et al., 2004;Nichol et al., 2006). Therefore, PRI in addition to being a water stressindicator, is also directly related to several physiological processesinvolved in the photosynthetic system.

The remote detection and monitoring of water stress is critical inmanyworld areaswherewater scarcity is amajor constraint to irrigatedagriculture, and is forcing farmers to reduce irrigation water use viadeficit irrigation (DI) (Fereres & Soriano, 2007). One of the DIapproaches is the regulated deficit irrigation (RDI),wherewater deficitsare imposed only during the crop developmental stages that are theleast sensitive to water stress (Chalmers et al., 1981). This practice wasoriginally proposed to control the vegetative vigour in high-densityorchards to reduce production costs and to improve fruit quality.However, it also saves irrigationwater, with the concomitant benefits ofreduced drainage losses (Fereres & Soriano, 2007). It has long beenknown that tree water deficits affect fruit quality parameters (Veih-meyer, 1927).However,whenwater deficits are imposed as inRDI, yieldand fruit size are not affected (Girona, 2002), while some qualityparameters such as total soluble sugars and total acidity increase(Crisosto et al., 1994; Girona et al., 2003; Mills et al., 1994). Theresponses to RDI are variable depending on the timing and severity ofwater deficits (Marsal & Girona, 1997; Girona et al., 2003) which varywithin a given orchard; thus the need for remote sensing tools thatcould assist in monitoring stress over entire orchards. Additionally, thechanges in irrigation depths with time and the lack of uniformity inwater application during the irrigation period emphasize the need for amethodology that would cover the entire season, integrating the short-term variations in tree water status. One option would be to use anintegrated measure over time of tree water status (Myers, 1988;Ginestar & Castel, 1996). González-Altozano and Castel (1999) relatedthe time integral of stem water potential with yield and fruit qualityparameters in citrus. Baeza et al. (2007) attempted the same approachon vineyards, finding a correlation between a water stress-integral andfinal berry size, although not with sugar composition. Although therelationships between water stress and fruit quality has been widelystudied, the conclusion is that there is a lack of reliable indicators thatpredictwith precisionfinal fruit quality, and therefore there is a need forfurther research concerning potential fruit quality indicators.

Remote sensing of fruit quality has been attempted by severalmeans such as by determining the vigour or total leaf area invineyards (Johnson et al., 2001, 2003; Lamb et al., 2004); by relatingquality parameters in water-stressed mandarin trees to spectralchanges in the red and green channels (Kriston-Vizi et al., 2008), andby using high spatial resolution airborne thermal imagery to outlinerelationships of olive fruit size, weight, and oil content againstthermal water stress indicators (Sepulcre-Cantó et al., 2007).

In this work, the PRI has been used to assess fruit quality parametersin peach and orange orchards under variouswater regimes. A time-seriesof airborne PRI imagery over a peach and an orange orchard underdifferent irrigation treatments were acquired and related to fruit qualityat harvest. Furthermore, a 3D radiative transfer model was used to assessthe applicability of this method to medium resolution PRI imageryfor extended monitoring of crops at larger scales. For this purpose,simulations using different soil backgrounds were conducted and theoutput spectral informationwas evaluated at different spatial resolutions.

2. Material and methods

2.1. Study sites

The experimental areas are located in Western Andalucía, Spain, aregion of Mediterranean climate characterized by warm and dry

summers and cool and wet winters, with an average annual rainfall ofover 550 mm.

The first study site was located on a commercial peach orchardplanted in 1990 in a 5×3.3 m grid on a deep soil withmoderately highwater holding capacity and classified as Typic Xerofluvents inCordoba, Spain (37.5°N, 4.9°W) (Fig. 1a). Two experiments werecarried out in this location. One experiment was conducted using a setof eight rows of nine peach trees (Prunus persica cv. “BabyGold 8”).Within that set, 18 trees were drip irrigated starting in mid May (endof Stage I of fruit growth and beginning of Stage II) at a rate that metthe evapotranspiration (ET) requirements (full-irrigated treatment orcontrol). Additionally, three different RDI treatments were applied toplots of 12 trees each, by varying the onset of re-irrigation as Stage IIIof fruit growth commenced, following uniformwater deficits imposedin Stage II. The dates of onset of re-irrigation (at 160% of ET) were 4July, 11 July and 17 July, respectively. In the same commercial orchard,another experiment was conducted on nectarine trees (Prunus persicacv. “Sweet Lady”). Five rows of 30 trees each were irrigated meetingthe ET requirement (following the commercial orchard schedule),while another six rows of 30 trees each were subjected to an RDIregime that imposed water deficits until 30 June, the beginning ofStage III (Fig. 1b).

The second study site was located near Seville, Southern Spain(37°N, 5.7°W), on an 82 ha commercial citrus orchard (Fig. 1c showsthe subset where the experiment was established). The light-texturedsoil is classified as Rodoxeralf, with an approximate depth of 3 m. Thetrees were planted in 1997 in a 7×3 m pattern on a N–S orientation.The experiment was a randomized block design with six replications,each individual plot composed of five rows of three orange trees(Citrus sinensis L. cv. “Navelina”). Three different drip irrigationtreatments were applied:, i) the control, that followed the orchardschedule which is designed to meet ET requirements for maximumproduction; ii) an over-irrigated treatment that applied 137% ofcontrol; and iii) a DI treatment that applied 62% of control. Table 1 liststhe treatments, irrigation periods and depths, harvest and imageryacquisition dates for each of the study sites.

2.2. Field data

From the beginning of June till harvest, stomatal conductance (G)and stem water potential (Ψ) were measured weekly at midday witha leaf porometer (model SC-1, Decagon Devices, Washington, DC,USA) and a pressure chamber (PWSC Model 3000, Soil MoistureEquipment Corp., California), respectively. For the first experimentalsite, measuredΨ values for the three RDI treatments were normalizedby dividing them into the control Ψ values of the full-irrigatedtreatment.

Leaf spectral measurements of adaxial surfaces were taken with aleaf probe that was attached to a field spectrometer ASDmeasuring inthe spectral range of 400–1100 nmwith a 25° field of view (FieldSpecHandheld Pro, ASD Inc., CO, USA). Measurements were taken on fourleaves per crown, which were collected and immediately frozen inliquid nitrogen. For these, leaf spectral properties and indices werecalculated to assess relationships with pigment concentrationsdetermined after extraction through chromatography (see below).

Two square centimeters of leaf tissue were obtained with a corkborer from each of the four leaves per tree crown. The discs werefrozen in liquid nitrogen in the field and later kept under −20 °C inmicrofuge tubes. Each leaf disc set corresponding to each crown wasground in a mortar on ice with liquid nitrogen and acetone (in thepresence of Na ascorbate) up to 5ml. Then, the extract was filteredthrough a 0.45 μm filter to separate vegetation residues and Naascorbate and left in dark tubes at −20 °C for pigment analysis, asreported by Abadía and Abadía (1993). First, absorption at 470, 644.8and 661.6 nm was measured with a spectrophotometer in order toderive chlorophyll a and b, and total carotenoid concentrations as

Fig. 1. Overview of the experiments conducted on peach trees (a); nectarines (b); and orange trees (c). The irrigation treatments for the peach study area consisted of fully-irrigatedand three deficit irrigation treatments (RDI1, RDI2 and RDI3). The nectarine field was divided into one fully-irrigated treatment and five rows of trees under regulated deficitirrigation (RDI). In the orange tree experiment there were three different irrigation treatments: over-irrigated, control and deficit-irrigated.

288 L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

described by Abadía and Abadía (1993). Samples were injected in a100×8 mm Waters Novapak C18 radial compression column (4 µmparticle size) with a 20 µl loop and mobile phases were pumped by aWaters M45 high pressure pump at a flow of 1.7 ml/min (Larbi et al.,2004).

In all three orchards, between 12 and 36 trees located in the centreof the irrigation treatments were monitored and harvested individ-ually. All fruits from each tree were immediately weighted and theirdiameters measured. Later, eight fruits per tree were selected

randomly for a physiochemical and organoleptic characterization ofTotal Soluble Solids (TSS) and Total Acidity (TA) used to calculate theratio (TSS/TA). The combination of TSS and TA in the ratio TSS/TA is anindicator of both sweetness and fruit acidity, giving more informationthan TSS or TA separately (Crisosto et al., 2006). Fruit juice wasobtained, filtered and measured with a pH meter (pH-Burette 24,Crison, Spain). To measure Titratable Acidity, 6ml of juice from eachsample were mixed with 50 ml of water and used for titrating with0.1 N NaOH to an end point of pH 8.2. The total volume of NaOH is

Table 1Summary of each study site, harvest, and imagery acquisition dates.

Species Treatments Irrigation strategy Withheld period Irrigation dose Harvest Flights

Nectarine Control Sustained 100% ET From 07/15 to 08/07 06/12, 06/19, 06/25, 07/01, 07/04, 07/17, 07/23, 07/29,08/07, 08/12RDI Regulated 05/21 to 06/30 Re-watering:200% ET, later:

100% ETPeach Control Sustained 80% ET

RDI-1 Regulated 05/19 to 07/04 Re-watering:160% ET, later:80% ET

From 08/06 to 08/28 06/19, 06/25, 07/01, 07/04, 07/17, 07/23, 07/29, 08/07,08/12, 08/21

RDI-2 Regulated 05/19 to 07/11 Re-watering:160% ET, later:80% ET

RDI-3 Regulated 05/19 to 07/18 Re-watering:160% ET, later:80% ET

Orange Control Sustained 100–130% ET 11/03 09/16Deficit Sustained 62% ETOver-irrigated

Sustained 137% ET

The treatments under RDI had a “withheld period” (no irrigation) followed by a re-watering period until their stem water potential were equal to the control.

289L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

measured and used to calculate the Titratable Acidity using theEq. (1).

% Acid =mlðNaOHÞ used × ð0:1NNaOHÞ × ðmilliequivalent factorÞ × 100

grams of sampleð1Þ

The milliequivalent factor is a coefficient dependant on the mostpredominant acid in the fruit. In the case of peach trees, malic acid ispredominant and the milliequivalent factor is 0.067. For orange trees,citric acid is predominant, with a milliequivalent factor of 0.064.Soluble solids concentration (SSC%, °Brix) was determined in a smallsample of fruit juice using a hand-held refractometer (Atago, ATC-1E,Japan).

Pigment concentrations were derived from the total area of thepeaks in the chromatogram using previously determined calibrationcoefficients calculated by injecting pure pigments into the HPLCcircuit. From the xanthophyll pigment concentrations of violaxanthin(V), antheraxanthin (A), and zeaxanthin (Z), the epoxidation state(EPS) was calculated as (V+0.5⁎A) /(V+A+Z) (Thayer & Björkman,1990). The EPS was calculated from the leaf set corresponding to eachcrown and compared with the PRI calculated as (R570−R531)/(R570+R531) (Gamon et al., 1992), from the spectra measured in thefield on the same four leaves sampled.

2.3. Airborne imagery

A 6-band multispectral camera (MCA-6, Tetracam, Inc., California,USA) was flown in the summer of 2008 at 150 m above the groundlevel using an unmanned aerial vehicle (UAV) (Berni et al., 2009),acquiring a time-series of 12 images on different dates from the twostudy sites. Eleven images were acquired on the commercial fieldcovering the nectarine and peach experiments from the beginning ofthe Stage II of fruit growth (12th of June) till the end of harvest (21August). In the orange orchard, the water status of the differentirrigation treatments was kept constant. Hence, there was no need forcharacterizing the water stress over the whole fruit growing period,and a single image acquired on 16 September was used. The camera

Table 2Overview of the vegetation indices used in this study and their formulation, with Rx being

Normalized Difference Vegetation IndexTransformed Chlorophyll Absorption in Reflectance Index/Optimized Soil-AdjustedVegetation Index

Simple RatioPhotochemical Reflectance Index

has six image sensors with 25 mm diameter bandpass filters of 10 nmFWHM (Andover Corporation, NH, USA). The image resolution is1280×1024pixels with 10-bit radiometric resolution and optical focallength of 8.5 mm, yielding a ground-based spatial resolution of 15 cmat 150 m altitude. The bandsets used in each of the study sitesincluded those centered at 530 and 570 nm used to calculate the PRIindex, as well as 550, 670, 700 and 800 nm to calculate the TCARI/OSAVI index for chlorophyll content estimation (Haboudane et al.,2002), and the NDVI (Rouse et al., 1974), and the SR for LAIestimation. An overview of the spectral indices used in this studycan be found in Table 2. Geometric calibration of airborne data wasconducted as explained in Berni et al. (2009). The camera wasradiometrically calibrated using coefficients derived from measure-ments made with a uniform calibration body (integrating sphere,CSTM-USS-2000C Uniform Source System, LabSphere, NH, USA) atfour different levels of illumination and six different integration times.Radiance values were later converted to reflectance using totalincoming irradiance simulated using sunphotometer (Microtops,Solar Light inc.) data taken in the field at the time of imageryacquisition.

A thermal camera (Thermovision A40M; FLIR, USA) was installedonboard the airborne platform. Its image resolution was320×240pixels and 16bits of at-sensor calibrated radiance with a40° FOV lens, yielding 40 cm spatial resolution at 150 m altitude. Theimage sensor is a Focal Plane Array (FPA) based on uncooledmicrobolometers with a spectral range of 7.5–13 μm, yieldingcalibrated radiance in the range of 233–393 K. The methodology forobtaining surface temperature from radiance temperature by remov-ing atmospheric effects using a single-channel atmospheric correctionis explained in Berni et al. (2009).

As PRI is an index related to light use efficiency, the index wasnormalized with the incoming PAR over the time-series at the time ofimage acquisition, calculated as the integral of the irradiance in thevisible region of the electromagnetic spectrum (range 400–700 nm).Thus, PRI/PAR values could be used to study the time-series over thestages II and III of fruit growth. Reflectance values were obtained forthe six spectral bands for every single crown for the whole time-series. The high spatial resolution allowed the identification of each

the reflectance at xnm.

NDVI = R800−R670R800 + R670

Rouse et al. (1974)TCARI =OSAVI = 3⁎½ðR700−R670Þ−0:2⁎ðR700−R550 Þ⁎ðR700 = R670Þ�

ð1 + 0:16Þ⁎ðR800−R670 Þ = ðR800 + R670 + 0:16Þ Haboudane et al., 2002

SR = R800R670

Asrar et al. (1985)

PRI = R570−R531R570 + R531

Gamon et al. (1992)

290 L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

individual tree and the possibility of selecting pure sunlit vegetationpixels. The pixels at the edge of the crowns were not used to extractthe spectral information. Additionally, individual crown temperature(Tc) was extracted from thermal imagery avoiding crown edge pixelsand was normalized with air temperature (Ta) to use the value ofTc−Ta for the analysis. The integral over time was calculated for fiveindices: PRI/PAR, Tc−Ta, NDVI, SR, and TCARI/OSAVI for the periodfrom 19th of July to 7th of August when the harvest started. Forindices having both negative and positive values, the minimum valuefound for each index along the whole period was used as a referenceline to calculate the integral for all the trees, as explained in Myers(1988).

In order to study the effect of the spatial resolution on PRI, theaggregated reflectance including shadows, soil and crown compo-nents was extracted for each crown in all images acquired, and theintegral calculated. For this purpose, regions of interest comprisingeach individual crown, and the adjacent soil were created and thespectral information extracted.

2.4. Radiative transfer modelling

The 3D Forest Light interaction radiative transfer model (FLIGHT,North, 1996) was used to investigate the application of thismethodology to larger spatial resolution imagery, by assessing canopyaggregation effects on PRI. The FLIGHT model has been successfullyused to simulate PRI in previous studies (Barton and North, 2001;Suárez et al., 2008, 2009). In this case, typical peach leaf spectralcharacteristics and tree structural parameters were inputs forsimulating vegetation cover ranging from 10% to 100% on threedifferent soil types. Typical peach leaf reflectance and transmittance

Fig. 2. Spectra from three soil types were used as input for radiative transfer modelling ofvegetation coverage (b, c and d, respectively).

spectra were simulated using PROSPECT model (Jacquemoud andBaret, 1990) for N=1.6, Cab=40 μg/cm2, Cm=Cw=0.015 andCs=0. The input values of N, Cm, Cw, and Cs were found in previousliterature on peach trees (Suárez et al., 2009), the input value for Cabwas defined as the average of the chlorophyll concentrationsestimated by destructive methods. The canopy input parametersused were crown LAI=2.5, LAD=spherical and leaf size=0.02 m.The solar geometry corresponded to mid July at 10:00 GMT inCordoba, Spain (solar zenith=23° and solar azimuth=124°). Fig. 2represents the three generic soil spectra used (Fig. 2a) and threesimulations conducted for dark soil and 30% (Fig. 2b), 50% (Fig. 2c),and 70% vegetation cover (Fig. 2d). From each simulation, crownreflectance and scene reflectance were extracted digitalizing regionsof interest on the simulated scene image to calculate crown PRI andscene PRI, respectively. Crown PRI values were compared to scene PRIvalues in order to assess the effects caused by the spatial resolution onthe index. One of the outputs of FLIGHT radiative transfer model is thepercentage of each element: shadowed soil, sunlit soil, shadowedvegetation and sunlit vegetation in the whole simulated scene. Thosepercentages were used to assess the magnitude of the error whenderiving EPS from an aggregated pixel of soil, shadows and vegetationas a function of the soil type. For this purpose, scenes with avegetation cover of 50% were simulated using the soil spectra in Fig. 2.The aggregated pixel reflectance was calculated as the sum of thefractional covers of each element in the scene multiplied by itsreflectance. The vegetation fractional cover was considered to be thesum of the fractional covers for sunlit and shadowed vegetation fromthe FLIGHT output. The aggregated pixel reflectance was calculated asthe sum of the fractional covers of each element in the scenemultiplied by its reflectance. The vegetation fractional cover was

vegetation scenes. (a). Simulations were conducted for dark soil at 30%, 50% and 70%

291L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

considered to be the sum of the fractional covers for sunlit andshadowed vegetation from the FLIGHT output. The aggregated scenereflectance was the result of using tree reflectance which corre-sponded to the vegetation fractional cover in the scene. Threedifferent scenes were recreated per tree using three soil types. Foreach tree, the EPS value was compared to the aggregated PRI

Fig. 3. (a) Relationship between the epoxidation state (EPS) calculated fromxanthophyll pigment extraction and PRI from spectral measurements on the sameleaves from a total of 12 tree crowns. Relationship between the EPS calculated from theaveraged xanthophyll content of four leaves per crown and spectral indices extractedfrom imagery for the crowns: PRI index (b) and NDVI (c).

Fig. 4. Correlation between pre-harvest EPS in the nectarine orchard experiment andthe fruit quality ratio, Total Soluble Solids/Tritatable Acidity (TSS/TA).

calculated from tree reflectance and the reflectance of the fractionsof shadowed and sunlit soil.

3. Results and discussion

Fig. 3a shows that, at the leaf level, the EPS calculated frompigment determination methods was well correlated with leaf PRIcalculated from the same leaves collected in the field. Leaves withhigher EPS values, corresponding to a high concentration of thephotosynthetic active pigment violaxanthin over the whole xantho-phyll pool, and consequently less stressed, presented lower PRIvalues. Lower values of PRI are the consequence of lower absorption at530 nm using the presented formulation of PRI corresponding to alower concentration zeaxanthin, which is the xanthophyll pigmentinhibiting photosynthesis (Gamon et al., 1992). At the crown level, thephoton interaction with vegetation structure and the effect of leafangle distribution result in lower PRI values as we can see in Fig. 3band c. These figures depict the relationships obtained between EPS,crown PRI (r2=0.41), and crown NDVI (r2=0.15) extracted from thecanopy images. The data shown on Fig. 3b and c suggest that PRIchanges are driven by physiological processes related to xanthophyllpigments and not by vegetation structure. At both leaf and canopyscales, PRI and chlorophyll a+b content were not correlated, yieldingcoefficients of determination of r2=0.04 and r2=0.01 respectively(data not shown), indicating that PRI changes were not driven bychlorophyll content differences. Several authors have previouslyfound an inverse relationship between EPS and PRI (Filella et al.,

Table 3Coefficients of determination (r2) in the orange experiment, between indices [PRI, T,NDVI, SR and TCARI/OSAVI] and post-harvest fruit quality parameters [TSS, TotalSoluble Solids; TA, Tritatable Acidity; the ratio, TSS/TA; and the median of the fruit size].

Orange

TSS TA TSS/TA

Fruitsize

Indices for water stress detection PRI 0.17 0.50⁎⁎ 0.50⁎⁎ 0.11T (K) 0.00 0.01 0.00 0.47⁎

Structural indices NDVI 0.28⁎ 0.33⁎ 0.16 0.17⁎

SR 0.28⁎ 0.34⁎ 0.17 0.18⁎

Chlorophyll index TCARI/OSAVI

0.24⁎ 0.35 0.28 0.32

Pearson's analysis.⁎⁎Correlation is significant at the 0.01 level (2-tailed) (In bold).⁎Correlation is significant at the 0.05 level (2-tailed).

Fig. 5. Relationship between the ratio, Total Soluble Solids/Tritatable Acidity (TSS/TA) and: (a) crown PRI; (b) canopy T; (c) NDVI; and (d) TCARI/OSAVI extracted from imagery onthe orange orchard study area.

292 L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

1996; Gamon et al., 1992, 1997; Guo et al., 2006; Nichol et al., 2006),as shown in Fig. 3b. The results in Fig. 3 demonstrate the link betweenPRI and the xanthophyll cycle in peach at both leaf and canopy scales.

Fig. 6. Overview of the time-series for the peach deficit irrigation treatmentsnormalized by the fully-irrigated treatment values of PRI/PAR and stemwater potential(SWP) from the beginning of Stage II of fruit growth until harvest. I1, I2 and I3correspond to the dates in which RDI1, RDI2 and RDI3 treatments were re-irrigated torecover from water stress.

The different irrigation treatments generated variability in fruitquality, as assessed by the TSS/TA ratiomeasured at harvest. The TSS/TAvariability in the peach orchard with four irrigation treatments wasgreater than the TSS/TA variability found in the nectarine and orangeorchards, where only one deficit irrigation treatmentwas applied. Fig. 4shows the relationship between EPS at crown scale and TSS/TA

Table 4Coefficients of determination (r2) for peach and nectarine between the time-seriesintegral of indices [PRI/PAR, Tc−Ta, NDVI, SR and TCARI/OSAVI] and the fruit qualityparameters [TSS, Total Soluble Solids; TA, Tritatable Acidity; the ratio, TSS/TA and themedian of the fruit size].

Peach Nectarines

TSS TA TSS/TA

Fruitsize

TSS TA TSS/TA

Fruitsize

Indices forwaterstressdetection

∫(PRI/PAR)dt

0.28 0.48⁎ 0.72⁎⁎ 0.01 0.05 0.22 0.61⁎⁎ 0.32⁎

∫(Tc−Ta)dt

0.01 0.27 0.21 0.27 0.04 0.02 0.24 0.28

Structuralindices

∫(NDVI)dt 0.60⁎ 0.00 0.55⁎⁎ 0.00 0.07 0.13 0.16 0.46∫(SR)dt 0.32 0.36⁎ 0.62⁎⁎ 0.00 0.15 0.03 0.03 0.13

Chlorophyllindex

∫(TCARI/OSAVI)dt

0.55⁎⁎ 0.05 0.38⁎ 0.05 0.02 0.08 0.26 0.28

Pearson's analysis.⁎⁎Correlation is significant at the 0.01 level (2-tailed) (In bold).⁎Correlation is significant at the 0.05 level (2-tailed).

Fig. 7. Integral of PRI/PAR and Tc−Ta from imagery for Stages II and III of fruit growth versus the fruit quality ratio, TSS/TA for peach (a and c) and nectarine (b and d).

293L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

measured at harvest for the nectarine experiment. The higher values ofEPS, obtained from trees under full irrigation, corresponded to lowervalues of TSS/TA found in fruits of that treatment. Previous studies onpeach have demonstrated the inverse relationship between tree waterstatus at Stage II of fruit growth and the ratio TSS/TA (calculations fromdata published by Crisosto et al., 1994; Gelly et al., 2003).

In the orange orchard experiment, one treatment was subjected tosustained but mild deficit irrigation throughout the irrigation season.Measurements of PRI, NDVI, SR, TCARI/OSAVI and temperatureextracted from imagery were related to fruit parameters. Therelationships of fruit quality parameters (TSS, TA, TSS/TA and fruitsize) with the physiological and structural indices are listed in Table 3.PRI showed the highest correlation with TA and TSS/TA, which areconsidered important indicators of fruit quality. On the contrary,crown temperature was not associated with any of the qualityparameters, with one exception (fruit size; r2=0.47). This isconsistent with the stem water potential measurements, whichindicated no significant differences between the deficit irrigationtreatment and the control. Consequently, crown temperature differ-ences between the two treatments were hardly detectable. The PRIimage on the orange orchard was acquired during the stage of rapid

fruit growth. Fig. 5 shows the relationship of TSS/TA with: a) PRI; b)temperature; c) NDVI; and, d) TCARI/OSAVI. Crown PRI correlatedreasonably well with TSS/TA (r2=0.50), while other indicators ofwater stress, such as temperature, structural indices such as NDVI andSR, and an index related to chlorophyll content TCARI/OSAVI were notrelated to TSS/TA, (coefficients of determination of 0, 0.16, 0.17 and0.18, respectively; Fig. 5 and Table 3). Although canopy temperature isa reliable remote sensing water stress indicator (Jackson et al., 1977),crown temperature values should be related to the instantaneoustranspiration rate at the time of image acquisition. By contrast, crownPRI values reflect physiological processes related to photosynthesis asaffected by water stress. Because fruit quality is more tied tophotosynthesis and carbonmetabolism, PRI may be a better estimatorof fruit quality than other established water stress indicators that arerelated directly to transpiration such as crown temperature.

In the peach experiment, the time courses of water deficits andrecovery were monitored by measuring crown PRI from imagery andSWP in the field. The water-stressed trees showed lower SWP andhigher crown PRI values (representing stress conditions) than thecontrol trees. Fig. 6 shows the time-series of PRI/PAR and thenormalized SWP values (values divided by those of the control

Fig. 8. (a) The peach orchard showing four trees corresponding to four irrigation treatments: full irrigation, RDI1, RDI2 and RDI3; (b to e) zoom on selected trees showing TSS/TAvalues; (f to i) zoom on selected and surrounding trees showing the integral of PRI/PAR.

294 L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

treatment). During the water deficit period that corresponded toStage II of fruit growth, the values of PRI/PAR for the three deficitirrigation treatments were similar and higher than the control values,indicating water stress (Fig. 6). Similarly, the SWP of those treatmentswere lower than that of the fully-irrigated treatment. As the rapidgrowth or stage III of fruit growth began in 4 July (“Irrigation 1”), theRDI-1 treatment was re-irrigated until its SWP recovered to controlvalues, a process that took 5days. One week later (“Irrigation 2”), theRDI-2 treatment was re-irrigated, and the PRI/PAR for both RDI-1 andRDI-2 treatments fell below the control values (Fig. 6) showingrecovery of water status. Equally, SWP values recovered in RDI-1 andRDI-2 until they reached the values of the fully-irrigated treatment,although the RDI-2 treatment took ten days to recover (Fig. 6). TheRDI-3 treatment entered the recovery phase at “Irrigation 3”; both theSWP and PRI/PAR records showed that recovery in this treatment

occurred after 13 days, taking longer to recover from the more severewater stress than the other two RDI treatments. The results of Fig. 6clearly demonstrate that the PRI/PAR measurements tracked theevolution of tree water status in the various treatments.

In the peach experiment, the variation in irrigation regimes withtime indicated that a single measurement of PRI could not glean stresshistory. Therefore, it was necessary to use a time-series of dataacquired during fruit growth to accurately describe the differenttreatments. Table 4 presents the results of analyzing the indices time-series (computed over Stages II and III of peach and nectarine fruitgrowth) against the parameters TSS, TA, TSS/TA and fruit size. Again,the ratio TSS/TA was best correlated with the integral of PRI/PAR forpeach and nectarine trees. Fig. 7 shows an independent analysis of thetwo water stress indicators: PRI/PAR and Tc−Ta. Fig. 7a and c showthe relationship of the integral of PRI/PAR and Tc−Ta with TSS/TA for

Fig. 9. (a) Zoom on an image from the peach study area; (b and c) area corresponding toaggregated crown, soil and shadows; (d) high spatial resolution imagery enablingwithin-crown separation of sunlit (yellow) and shaded vegetation (red).

295L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

peach, and Fig. 7b and d show the same for nectarines. The correlationof TSS/TA with the integral of PRI/PAR calculated for every singlecrown using a set of ten images taken in different days yielded linearrelationships of r2=0.72 for the peach tree experiment, and r2=0.61for the nectarine orchard (Fig. 7a and b, respectively). By contrast,relationships between the integral of Tc−Ta with TSS/TA, yieldedmuch lower linear relationships for peach (r2=0.21) and nectarines(r2=0.25) (Fig. 7c and d). A possible explanation for the difference inbehaviour between PRI and Tc−Ta lies in the changes in carbonpartitioning in response to mild water deficits. It has been shown thatdeficit irrigation alters the distribution of carbon, increasing theallocation to fruits (Fereres and Soriano, 2007). Mild water deficitsthat hardly would affect transpiration (and hence Tc−Ta) may havesome effects on carbon metabolism that are reflected in the lightreactions of photosynthesis and are detected by PRI measurements.

The effect of water stress on fruit size, an important commercialparameter (Guardiola & García-Luis, 2000), was studied in peach andnectarine using multispectral imagery acquired during Stage III ofexponential fruit growth, when the absence of water stress is critical(Génard & Huguet, 1996). PRI/PAR time-series computed over StageIII of fruit growth correlated well with fruit size at harvest (r2=0.51,data not shown), while the correlation of Tc−Ta time-series yielded amuch smaller coefficient of determination (r2=0.15). Fig. 8 presentsan overview of the peach experiment with the different treatmentsand a zoom on four of the monitored trees, each of themcorresponding to a different irrigation schedule: Full-irrigated, RDI-1, RDI-2 and RDI-3 (Fig. 8a). Fig. 8b to e show the ratio TSS/TA for thefour trees. In Fig. 8f to i, the integral of PRI/PAR is shown for each of theselected and surrounding trees, demonstrating that the variabilityamong trees of the water stress-integral within-treatments may bedetected using an image-based methodology. The values of theintegral of PRI/PAR and TSS/TA represented using the same color code,appear the same for the four selected trees, demonstrating that thedetection of fruit quality is possible using the time-series of PRI/PAR.Moreover, the use of remote sensing for the assessment of fruit qualityparameters permits the spatial characterization of an entire orchard.

The assessment of the influence of the imagery spatial resolutionon PRI was studied using image data and radiative transfer modelling.The performance of the integral of PRI/PAR for different spatialresolutions was assessed using data from imagery. First, only thereflectance extracted from pure crowns was used (Fig. 9a), and thenthe reflectance extracted from aggregated pixels, including soil,shadows and crowns (Fig. 9b). The use of high spatial resolutionimagery (zoom shown in Fig. 9c) allowed the classification of thecrown into sunlit vegetation pixels and shadowed vegetation pixels.Fig. 9d presents an example of a supervised maximum likelihoodclassification in which within-crown sunlit vegetation and shadowscan be identified. When the integral of PRI/PAR is calculated usingpure crown reflectance (Fig. 10), the integral at crown scale, as afunction of the EPS along the fruit growth, is well correlated with theTSS/TA ratio (r2=0.72, Fig. 10a). The integral of PRI/PAR extractedfrom pure crown pixels (i.e., high spatial resolution) versus theintegral of PRI/PAR extracted from aggregated pixels where there iscrown, soil and shadows (low-spatial resolution) yielded a very lowassociation (r2=0.06, Fig. 10b). The lack of relationship when scenecomponents are aggregated in a mixed pixel suggests that the PRI vs.EPS relationship is lost when pure sunlit crowns are not selected. Infact, the relationship of the integral of “low-spatial resolution” PRIwith TSS/TA yielded a coefficient of determination of 0.25 (Fig. 10c),versus the r2=0.72 when using high spatial resolution PRI. Theresults of modelling PRI for pure crown and aggregated pixels suggesta large influence of soil and shadows on the index (Fig. 11), whichprevents detecting the physiological responses to mild water stresslevels. Aggregated PRI values extracted from the full reflectance scenediffer greatly from pure crown PRI values as vegetation coveragedecreases, increasing the proportion of soil and shadowed back-

ground in the aggregated pixel from 100% to 10% vegetation cover.The effects on the index as a function of three soil types (spectra fromFig. 2) were assessed. Dark soils showed a higher influence onaggregated PRI values (see the separation from the 1:1 line), even forhigh percentage vegetation coverage (Fig. 11). The dark soil spectrumis low and flat in comparison with the two other soil spectra used,then the influence on PRI is higher than the influence of soils withsteeper spectra. The possible errors using low-resolution imagery arepresented in Fig. 12. The error bars in Fig. 12b represent the variabilityof the PRI value for an aggregated pixel depending on the soil type.These modelling results emphasize the need for high spatialresolution imagery to reproduce the results obtained in this study,as medium resolution PRI imagery would be heavily affected bybackground and shadow components, greatly reducing the sensitivityto physiological indicators of stress, such as PRI. Future developmentsin this field may focus on minimizing background effects in order touse our methodology with lower spatial resolution imagery. Oneapproach could be the use of a soil-adjusted vegetation index to dealwith variable backgrounds. The first soil-adjusted vegetation indexpresented was SAVI (Huete, 1988). Subsequently, additional indiceshave been proposed such as TSAVI (Baret & Guyot, 1991), MSAVI (Qiet al., 1994), OSAVI (Rondeaux et al., 1996), and GESAVI (Gillabertet al., 2002), among others. Other authors have successfullytransformed indices using soil-adjusted lines to minimize soilinfluence on spectral reflectance (Haboudane et al., 2002). Use of

Fig. 11. Pure crown PRI values extracted from radiative transfer simulation using the 3Dforest light interaction model (FLIGHT) versus aggregated PRI including crown, soil andshadows for vegetation cover ranging from 100% to 10% for three backgrounds. Leafinput parameters were N=1.6, Cab=40 μg/cm2, Cw=0.015, Cm=0.015 and Cs=0,canopy input parameters were LAD=spherical, crown LAI=2.5, solar zenith=23° andsolar azimuth=124°.

Fig. 12. (a) Relationship between EPS and pure crown PRI for peach trees; (b)relationship between EPS and PRI from aggregated pixels including different back-grounds. The error bars are function of the maximum and minimum scene PRIcorresponding to each EPS value.

Fig. 10. Relationship between the integral of PRI/PAR extracted from pure crowns andthe fruit quality ratio, TSS/TA (a); relationship between the integral of PRI/PARextracted from pure vegetation spectra vs the integral of PRI/PAR of aggregated crown,soil and shadows (b); relationship between the integral of PRI/PAR for aggregatedcrown, soil and shadows with TSS/TA (c).

296 L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

297L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

the above-mentioned approaches may adapt this methodology tolower spatial resolution imagery.

4. Conclusions

This study demonstrates the link between the epoxidation state ofthe xanthophyll cycle and the fruit quality measured in orchards underdifferent irrigation regimes, enabling the remote detection of fruitquality as a function of water stress using high-resolution airborne PRI.The PRI index measured at leaf scale was in agreement with theepoxidation state of the xanthophyll cycle calculated from destructivesampling. In addition, the airborne image-derived PRI values calculatedfrom pure crown reflectance were also in agreement with xanthophyllEPS measured on the same trees. Moreover, the time-series of airbornePRI normalized by the incoming PAR at the time of imagery acquisition(PRI/PAR) matched the changes in tree water status as affected by thedifferent irrigation regimes. In anorangeorchardunder sustaineddeficitirrigation, crown PRI calculated from the airborne imagery acquiredduring the fruit growth was well related to fruit quality. For peach andnectarine orchards subjected to period of water stress and recovery amethodologywas applied that used the time-series of PRI/PAR to assessthe fruit quality responses of the different stressed treatments.

Although the high-resolution PRI time-series exhibited a goodrelationship with fruit quality, crown airborne temperature acquiredover the same trees, the established method for remote sensing ofwater stress, did not yield comparable results and did not correlatewith fruit quality. This highlights the advantage of the PRI as a waterstress indicator related to other physiological processes and not onlyto transpiration, for the assessment of fruit quality. Finally, the use ofhigh spatial resolution imagery appears critical for a tight relationshipbetween PRI and EPS, and therefore for assessing fruit quality fromairborne PRI measurements. A radiative transfer simulation studydemonstrated the influence of soil and shadows on canopy reflectanceused to calculate PRI on aggregated pixels, suggesting the criticaleffects that soil variability has in the computation of the PRI index indiscontinuous orchard canopies. This work demonstrates the feasi-bility for assessing fruit quality in orchards using PRI when highspatial resolution remote sensing imagery is used (20 cm in thepresent study) as opposed to using lower spatial resolutions wherethe pixels represent a mixture of shadows, vegetation and soil. Thepractical implications of this approach to optimize harvest operationsand maximize revenues in horticultural crops based on fieldsegmentation as a function of fruit quality may be substantial.

Acknowledgements

Financial support from the Spanish Ministry of Science andInnovation (MCI) for the projects AGL2005-04049, EXPLORA-INGENIOAGL2006-26038-E/AGR, CONSOLIDER CSD2006-67, and AGL2003-01468, and from Gobierno de Aragón (group A03) is gratefullyacknowledged, and support in-kind provided by Bioiberica throughthe project PETRI PET2005-0616. Technical support from UAV Naviga-tion and Tetracam Inc. is also acknowledged. M. Medina, C. Ruz, R.Gutierrez, A. Vera, D. Notario, I. Calatrava and M. Ruiz Bernier areacknowledged for measurements and technical support in field andairborne campaigns. The Plant Stress Physiology Group of theExperimental Station of Aula Dei (CSIC) in Zaragoza is acknowledgedfor technical support on the leaf pigment extraction and quantification.

References

Abadía, A., & Abadía, J. (1993). Iron and plant pigments. In L. L. Barton & B. C. Hemming(Eds.), Iron chelation in plants and soil microorganisms (pp. 327−344). San Diego:Academic.

Asner, G. P., Carlson, K. M., &Martin, R. E. (2005). Substrate age and precipitation effectson Hawaiian forest canopies from spaceborne imaging spectroscopy. RemoteSensing of Environment, 98, 457−467.

Asrar, G., Kanemasu, E. T., Jackson, R. D., & Pinter, J. R. (1985). Estimation of total above-ground phytomass production using remotely sensed data. Remote Sensing ofEnvironment, 17, 211−220.

Baeza, P., Sánchez-de-Miguel, P., Centeno, A., Junquera, P., Linares, R., & Lissarrague, J. R.(2007). Water relations between leaf water potential, photosynthesis andagronomic vine response as a tool for establishing thresholds in irrigationscheduling. Scientia Horticulturae, 114, 151−158.

Baret, F., & Guyot, G. (1991). Potentials and limits of vegetation indices for LAI and PARassessment. Remote Sensing of Environment, 35, 161−173.

Barton, C. V. M., & North, P. R. J. (2001). Remote sensing of canopy light use efficiencyusing the photochemical reflectance index. Model and analysis. Remote Sensing ofEnvironment, 78(264), 273.

Berni, J. A. J., Zarco-Tejada, P. J., Suarez, L., & Fereres, E. (2009). Thermal and narrow-band multispectral remote sensing for vegetation monitoring from an unmannedaerial vehicle. IEEE Transactions on Geoscience and Remote Sensing, 47, 722−738.

Chalmers, D. J., Mitchell, P. D., & Van Heek, L. (1981). Control of peach tree growth andproductivity by regulated water supply, tree density and summer pruning. Journalof the American Society of Horticultural Sciences, 106, 307−312.

Cohen, Y., Alchanatis, V., Meron, M., Saranga, Y., & Tsipris, J. (2005). Estimation of leafpotential by thermal imagery and spatial analysis. Journal of Experimental Botany,56, 1843−1852.

Crisosto, C. H., Crisosto, G. M., Echeverria, G., & Puy, J. (2006). Segregation of peach andnectarine (Prunus persica (L.) Batsch) cultivars according to their organolepticcharacteristics. Postharvest Biology and Technology, 39, 10−18.

Crisosto, C. H., Johnson, R. S., Luza, J. G., & Crisosto, G. M. (1994). Irrigation regimes affectfruit soluble solids concentration and rate of water loss of ‘O'Henry’ peaches.Horticultural Sciences, 29, 1169−1171.

Dobrowsky, S. Z., Pushnik, J. C., Zarco-Tejada, P. J., & Ustin, S. L. (2005). Simple reflectanceindices track heat and water stress-induced changes in steady-state chlorophyllfluorescence at the canopy scale. Remote Sensing of Environment, 97, 403−414.

Drolet, G. G., Huemmrich, K. F., Hall, F. G., Middleton, E. M., Black, T. A., Barr, A. G., &Margolis, H. A. (2005). A MODIS-derived photochemical reflectance index to detectinter-annual variations in the photosynthetic light-use efficiency of a borealdeciduous forest. Remote Sensing of Environment, 98, 212−224.

Evain, S., Flexas, J., & Moya, I. (2004). A new instrument for passive remote sensing: 2.Measurement of leaf and canopy reflectance changes at 531 nm and theirrelationship with photosynthesis and chlorophyll fluorescence. Remote Sensing ofEnvironment, 91, 175−185.

Fereres, E., & Soriano, M. (2007). Deficit irrigation for reducing agricultural water use.Journal of Experimental Botany, 58, 147−159.

Filella, I., Amaro, T., Araus, J. L., & Peñuelas, J. (1996). Relationship betweenphotosynthetic radiation-use efficiency of barley canopies and the photochemicalreflectance index (PRI). Physiologia Plantarum, 96, 211−216.

Fuentes, D. A., Gamon, J. A., Cheng, Y., Claudio, H. C., Qiu, H. L., Mao, Z., Sims, D. A.,Rahman, A. F., Oechel, W., & Luo, H. (2006). Mapping carbon and water vapourfluxes in a chaparral ecosystem using vegetation indices derived from AVIRIS. Re-mote Sensing of Environment, 103, 312−323.

Gamon, J. A., Peñuelas, J., & Field, C. B. (1992). A narrow-wave band spectral index thattrack diurnal changes in photosynthetic efficiency. Remote Sensing of Environment,41, 35−44.

Gamon, J. A., Serrano, L., & Surfus, J. S. (1997). The photochemical reflectance index: Anoptical indicator of photosynthetic radiation use efficiency across species,functional types, and nutrient levels. Oecologia, 112, 492−501.

Gelly, M., Recasens, I., Mata, M., Arbones, A., Rufat, J., Girona, J., & Marsal, J. (2003).Effects of water deficit during stage II of peach fruit development and postharveston fruit quality and ethylene production. Journal of Horticultural Science &Biotechnology, 78, 324−330.

Génard, M., & Huguet, J. G. (1996). Modeling the response of peach fruit growth towater stress. Tree Physiology, 16, 407−415.

Gillabert, M. A., González-Piqueras, J., García-Haro, F. J., & Meliá, J. (2002). A generalizedsoil-adjusted vegetation index. (2002). Remote Sensing of Environment, 82, 303−310.

Ginestar, C., & Castel, J. R. (1996). Response of young ‘Clementine’ citrus trees to waterstress during different phonological periods. Journal of Horticultural Science, 71(4),551−559.

Girona, J. (2002). Regulated deficit irrigation in peach. A global analysis. ActaHorticulturae, 592, 335−342.

Girona, J., Mata, M., Arbonès, A., Alegre, S., Rufat, J., & Marsal, J. (2003). Peach treeresponse to single and combined regulated deficit irrigation regimes under swallowsoils. Journal of the American Society of Horticultural Sciences, 128, 432−440.

González-Altozano, P., & Castel, J. R. (1999). Regulated déficit irrigation in “Clementinade Nules” citrus tree. I. Yield and fruit quality effects. Journal of Horticultural Scienceand Biotechnology, 74, 706−713.

Guardiola, J. L., & García-Luis, A. (2000). Increasing fruit size in citrus. Thinning andstimulation of fruit growth. Plant Growth Regulation, 31, 121−132.

Guo, J., & Trotter, C. M. (2004). Estimating photosynthetic light-use efficiency using thephotochemical reflectance index: Variations among species. Functional PlantBiology, 31, 255−565.

Guo, J., Trotter, C. M., & Newton, P. C. D. (2006). Initial observations of increasedrequirements for light-energy dissipation in ryegrass (Lolium perenne) whensource/sink ratios become high at a naturally grazed Free Air CO2 Enrichment(FACE) site. Functional Plant Biology, 33, 1045−1053.

Haboudane, D., Miller, J. R., Tremblay, N., Zarco-Tejada, P. J., & Dextraze, L. (2002).Integrated narrow-band vegetation indices for prediction of crop chlorophyll contentfor application to precision agriculture. Remote Sensing of Environment, 84, 416−426.

Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing ofEnvironment, 25, 295−309.

298 L. Suárez et al. / Remote Sensing of Environment 114 (2010) 286–298

Idso, S. B. (1982). Humidity measurement by Infrared Thermometry. (1982). RemoteSensing of Environment, 12, 87−91.

Idso, S. B. (1982). Non-water-stressed baselines: A key to measuring and interpretingplant water stress. Agricultural Meteorology, 27, 59−70.

Idso, S. B., Jackson, R. D., Pinter, P. J., Reginato, R. J., & Hatfield, J. L. (1981). Normalizingthe stress-degree-day parameter for environmental variability. Agricultural andForest Meteorology, 24, 45−55.

Jackson, R. D., Idso, S. B., Reginato, R. J., & Ehrier, W. L. (1977). Crop temperature revealsstress. Crop Soils, 29, 10−13.

Jackson, R. D., Idso, S. B., Reginato, R. J., & Pinter, P. J., Jr. (1981). Canopy temperature as acrop water stress indicator. Water Resources Research, 17, 1133−1138.

Jackson, R. D., Slater, P. N., & Pinter, P. J., Jr. (1983). Discrimination of growth and waterstress in wheat by various vegetation indices through clear and turbid atmo-spheres. Remote Sensing of Environment, 13, 187−208.

Jacquemoud, S., & Baret, F. (1990). PROSPECT: Amodel of leaf optical properties spectra.Remote Sensing of Environment, 34, 75−91.

Johnson, L. F., Bosch, D. F., Williams, D. C., & Lobitz, B. M. (2001). Remote sensing ofvineyard management zones: Implications for wine quality. Applied Engineering inAgriculture, 17, 557−560.

Johnson, L. F., Roczen, D. E., Youkhana, S. K., Nemani, R. R., & Bosch, D. F. (2003).Mapping vineyard leaf area with multispectral satellite imagery. Computers andElectronics in Agriculture, 38, 33−44.

Kriston-Vizi, J., Umeda, M., & Miyamoto, K. (2008). Assessment of the water stressstatus of mandarin and peach canopies using visible multispectral imagery. Bio-systems Engineering, 100, 338−345.

Lamb, D. W.,Weedom, M. M., & Ramley, R. G. V. (2004). Using remote sensing to predictgrape phenolics and colour at harvest in a Cabernet Sauvignon vineyard: Timingobservations against vine phenology and optimising image resolution. AustralianJournal of Grape and Wine Research, 10, 46−54.

Larbi, A., Abadía, A., Morales, F., & Abadía, J. (2004). Fe resupply to Fe-deficient sugarbeet plants leads to rapid changes in the violaxanthin cycle and otherphotosynthetic characteristics without significant de novo chlorophyll synthesis.Photosynthesis Research, 79, 59−69.

Leinonen, I., & Jones, H. G. (2004). Combining thermal and visible imagery forstimulating canopy temperature and identifying plant stress. Journal of Experimen-tal Botany, 55, 1423−1431.

Marsal, J., & Girona, J. (1997). Relationship between leaf water potential and gasexchange activity at different phonological stages and fruit loads in peach trees.Journal of the American Society of Horticultural Sciences, 122, 415−421.

Mills, T. M., Behboudian, M. H., & Tan, P. Y. (1994). Plant water status and fruit quality in“Braeburn” apples. Horticultural Science, 29, 1274−1278.

Myers, B. J. (1988). Water stress integral — A link between short-term stress and long-term growth. Tree Physiology, 4, 315−323.

Nakaji, T., Oguma, H., & Fujinuma, Y. (2006). Seasonal changes in the relationshipbetween photochemical reflectance index and photosynthetic light use efficiencyof Japanese larch needles. International Journal of Remote Sensing, 27, 493−509.

Nichol, C. J., Huemmrich, K. F., Black, T. A., Jarvis, P. G., Walthall, J. G., & Hall, F. G. (2000).Remote sensing of photosynthetic-light-use efficiency of boreal forest. Agriculturaland Forest Meteorology, 101, 131−142.

Nichol, C. J., Lloyd, J., Shibistova, O., Arneth, A., Röser, C., Knohl, A., Matsubara, S., &Grace, J. (2002). Remote sensing of photosynthetic-light-use-efficiency of aSiberian boreal forest. Tellus, 54B, 677−687.

Nichol, C. J., Rascher, U., Matsubara, S., & Osmond, B. (2006). Assessing photosyntheticefficiency in an experimental mangrove canopy using remote sensing andchlorophyll fluorescence. Trees, 20, 9−15.

North, P. R. J. (1996). Three-dimensional forest light interactionmodel using aMontecarlomethod. IEEE Transactions on Geosciences and Remote Sensing, 34, 946−956.

Peguero-Pina, J. J., Morales, F., Flexas, J., Gil-Pelegrín, E., &Moya, I. (2008). Photochemistry,remotely sensed physiological reflectance index and de-epoxidation state ofxanthophyll cycle in Quercus coccifera under intense drought. Oecologia, 156, 1−11.

Qi, J., Chehbouni, Al, Huete, A. R., Kerr, Y. H., & Sorooshian, S. (1994). A modified soiladjusted vegetation index (MSAVI). Remote Sensing of Environment, 48, 119−126.

Rondeaux, G., Steven, M., & Baret, F. (1996). Optimization of soil-adjusted vegetationindices. Remote Sensing of Environment, 55, 95−107.

Rouse, J.W., Haas, R.H., Schell, J.A., Deering, D.W. & Harlan, J.C. (1974). Monitoring thevernal advancements and retrogradation of natural vegetation. in Nasa/Gsfc FinalReport (ed. MD, U.G.) p. 371.

Sepulcre-Cantó, G., Zarco-Tejada, P. J., Jiménez-Muñoz, J. C., Sobrino, J. A., Soriano, M. A.,Fereres, E., Vega, V., & Pastor, M. (2007). Monitoring yield and fruit qualityparameters in open-canopy tree crops under water stress. Implications for ASTER.Remote Sensing of Environment, 107, 455−470.

Sepulcre-Cantó, G., Zarco-Tejada, P. J., Sobrino, J. A., Berni, J. A. J., Jiménez Muñoz, J. C., &Gastellu-Etchegorry, J. P. (2009). Detecting water status in open canopies withthermal ASTER imagery and DART radiative transfer simulation. Agricultural andForest Meteorology, 149, 962−975.

Serrano, L., & Peñuelas, J. (2005). Assessing forest structure and function from spectraltransmittance measurements: A case study in a Mediterranean Holm oak forest.Tree Physiology, 25, 67−74.

Sims, D. A., Luo, H., Hastings, S., Oechel, W. C., Rahman, A. F., & Gamon, J. A. (2006).Parallel adjustment in vegetation greenness and ecosystem CO2 exchange inresponse to drought in a Southern California chaparral ecosystem. Remote Sensingof Environment, 103, 289−303.

Strachan, I. B., Pattey, E., & Boisvert, J. B. (2002). Impact of nitrogen and environmentalconditions on corn as detected by hyperspectral reflectance. Remote Sensing ofEnvironment, 80, 213−224.

Suárez, L., Zarco-Tejada, P. J., Berni, J. A. J., González-Dugo, V., & Fereres, E. (2009).Modelling PRI for water stress detection using radiative transfer models. RemoteSensing of Environment, 113, 730−744.

Suárez, L., Zarco-Tejada, P. J., Sepulcre-Cantó, G., Pérez-Priego, O., Miller, J. R., Jiménez-Muñoz, J. C., & Sobrino, J. (2008). Assessing canopy PRI for water stress detectionwith diurnal airborne imagery. Remote Sensing of Environment, 112, 560−575.

Thayer, S. S., & Björkman, O. (1990). Leaf xanthophyll content and composition in sunand shade determined by HPLC. Photosynthesis Research, 23, 331−343.

Thenot, F., Méthy, M., & Winkel, T. (2002). The Photochemical Reflectance Index (PRI)as a water-stress index. International Journal of Remote Sensing, 23, 5135−5139.

Trotter, G. M., Whitehead, D., & Pinkney, E. J. (2002). The photochemical reflectanceindex as a measure of photosynthetic light use efficiency for plants with varyingfoliar nitrogen contents. International Journal of Remote Sensing, 23, 1207−1212.

Veihmeyer, F. J. (1927). Some factors affecting the irrigation requirements of deciduousorchards. Hilgardia, 2, 125−284.